v0.7.0

This release improves progressive rendering, introduces various memory optimizations (especially in multithreaded mode), adds CMYK output support, adds support for capturing Exiv or other metadata boxes, and fixes various minor bugs.

What's Changed

• Increase cron cflite timeout to 3h. by @veluca93 in #881
• Configure rustfmt to format imports. by @veluca93 in #882
• Avoid making smooth unsqueeze steps access ModularBufferInfo. by @veluca93 in #883
• Use Jinc2 for smooth squeeze upscaling by @veluca93 in #885
• Update flake.lock by @tirr-c in #884
• Add MSRV and disable clippy lints with 1.98. by @veluca93 in #887
• Reduce memory usage for modular decoding. by @veluca93 in #886
• Remove support and use of internal image padding. by @veluca93 in #888
• Be more explicit on the requirements of JxlParallelRunner::run. by @veluca93 in #898
• Add CMYK interleaved pixel format by @hjanuschka in #891
• Premultiply expanded grayscale RGBA channels by @hjanuschka in #903
• Reject invalid blending reference frames by @hjanuschka in #902
• Make number of worker threads configurable in cli by @FooIbar in #904
• Extend the Zero-predictor fast path to scaled leaves by @hjanuschka in #906
• Update flake.lock to use Rust 1.98 by @tirr-c in #905
• Reject pixel format changes after the first frame header is decoded by @hjanuschka in #892
• Fix panic on grayscale non-XYB VarDCT frames by @Mrmaxmeier in #908
• Fix u8 overflow when a context map uses cluster index 255 by @Mrmaxmeier in #909
• Fix panic when a transform is applied to an empty palette channel by @Mrmaxmeier in #910
• Check the patches upsampling guard against the post-dim_shift value by @Mrmaxmeier in #913
• Fix panic in the Modular RLE fast path when LZ77 is disabled by @sl91994 in #912
• Improve smooth upscaling by @veluca93 in #911
• SIMD-accelerate AddNoiseStage by @veluca93 in #914
• 16bit coefficients for vardct by @veluca93 in #894
• Reduce memory usage and increase speed of noise generation. by @veluca93 in #915
• Fix nix flake check by adding dither_32x32.bin to source by @tirr-c in #918
• Fix various LLM-found minor bugs. by @veluca93 in #917
• Shared queue for transforms, and add JxlParallelRunner::run_ordered. by @veluca93 in #919
• Deallocate modular border buffers & full buffers in progressive decoding by @veluca93 in #899
• Fix whitepoint check by @veluca93 in #924
• Emit an error on image size layout overflow, instead of a panic. by @veluca93 in #925
• Reuse modular buffers (sharing with the pipeline) by @veluca93 in #920
• Better progressive decoding script by @veluca93 in #930
• Fix and parallelize LF preview rendering. by @veluca93 in #927
• Extract Exif box by @tirr-c in #926
• Support CMYK -> RGB conversion in jxl_cli by @FooIbar in #810
• Check the frame size limit against upsampled size. by @veluca93 in #932
• Fix check in consume_trailing_data for complete codestream by @tirr-c in #931
• Fix short reads when buffering container boxes by @gregbenz in #934
• Update flake.lock for Rust v1.98.1 by @tirr-c in #935
• Fix some box parser bugs by @tirr-c in #938
• Add additional checks to stop the large allocations regardless of sample_limit by @veluca93 in #936
• Avoid recapturing auxiliary boxes after frame seeks by @gregbenz in #937
• Add test aux_box_seek by @tirr-c in #939
• Use perfect best blue noise by @monad0 in #940
• Bump up SIMD/scalar comparison limits. by @veluca93 in #942
• Clean up and simplify the palette code. by @veluca93 in #943
• Add image crate decoding integration by @hjanuschka in #929
• 16-bit modular buffers by @veluca93 in #928
• Disable implicit cube/deltas for partial rendering. by @veluca93 in #945
• v0.7.0 by @veluca93 in #944

New Contributors

• @sl91994 made their first contribution in #912
• @gregbenz made their first contribution in #934
• @monad0 made their first contribution in #940

Full Changelog: v0.6.0...v0.7.0